#qi-hardware IRC log for Sunday, 2010-09-19

qwebirc93545has anyone figured how to boot the gui on debian for the nanonote?00:45
qi-bot[commit] Werner Almesberger: ben-baseframe-bottom-500um is done. http://qi-hw.com/p/ben-scans/1fd036e08:57
rafawpwrak: what is the problem with clocks?09:17
rafawpwrak: we have not touched nothing about .. so i do not have idea where to check if you need some info from jlime and clocks :)09:17
wpwrakrafa: good morning ! how's the head ? :) the problem is with the clocks gmenu2x in openwrt is messing with09:47
wpwrakrafa: at least in the latest "official" image, which is from june. so maybe this has changed since.09:47
mththe gmenu2x in git has been modified to use a cpufreq driver if one is present and otherwise it does not modify the clock frequencies09:59
kristianpaulwhz was modified?09:59
kristianpaulwhy09:59
mthbecause it's dangerous to poke into registers via /dev/mem09:59
kristianpaulhehe10:00
kristianpaulnoo ;)10:00
bartbeslol, so that's why there's a poke program now?10:00
mthpoking is useful for testing things, but it's not a long term solution10:01
kristianpaulwell in mobile/emdebbed stuff i bet is verz commmon10:01
rafawpwrak: ah.. okey.. well, we do not do anything like that10:01
mtha lot of bad practices are common in embedded ;)10:01
kristianpaullol10:02
kristianpaulmth: so zou think i should make a driver to drive io? and other stuff?10:02
kristianpaulmth: btw what was doing gmenu with /dev/mem what it needed?10:02
mthit was reprogramming the CPU and other clocks10:03
kristianpaulohh10:03
kristianpaulhe want we linux ;)10:03
kristianpaulwe/be10:03
kristianpaulahh the cloks !10:03
kristianpaulpower saving10:03
kristianpaulsound too?10:03
mththe sound clock is not derived from the PLL freq afaik10:04
kristianpaulok10:04
kristianpaulzear: are you aware off issues realted woth gmenu?10:04
kristianpaulwoth/with10:05
mthin general, if I/O is exclusively to one device and there is no need to handle interrupts and it is very device specific (no opportunity for creating an abstract interface) then it would be OK to do I/O directly from userland imo10:05
kristianpauli agree10:05
mthbut with the CPU clock there is a generic interface for that (cpufreq) and there is interaction with other drivers (many clocks derive from the PLL freq)10:05
kristianpaulindeed10:06
kristianpaulwhy we dont have dmenu in the nanonote?10:07
bartbesbecause I worked so hard on gmenu2x.. :P10:07
kristianpaul:)10:07
mthif you want to experiment, you can merge the cpufreq driver from my git tree: http://github.com/mthuurne/opendingux-kernel10:07
mthI'm not sure how much power is saved by lower frequencies though, since the CPU goes into idle mode when there is nothing to do10:08
mthon the Dingoo people use it for overclocking rather than underclocking10:08
kristianpaulbut but, what other drivers will required cpufreq in plaform/feature specific devices like the nanonote?10:09
mththe cpufreq driver is not mature yet, it can crash and likely will if you use the on-demand governor10:09
kristianpaulwell i bet in dingoo in is different bacuse the game oriented features10:09
wpwrakmth: (gmenu2x switch to cpufreq) ah, that's much better then, thanks.10:09
kristianpaulor let used to a 200Mhy nanonote ;)10:10
kristianpaulmhz*10:11
mthdmenu is nice if you have a few programs, but gmenu2x can handle larger numbers of programs well10:11
kristianpaulyou mean multi tasking?10:11
kristianpaulor sort of that...10:12
mthno, the number of programs displayed in the menu10:12
kristianpaulok i got the point10:12
kristianpaulhmm i remenerb last gmenu was able to add command line programs :)10:12
kristianpaulso bartbes is doing great ob :)10:13
kristianpauljob*10:13
kristianpaultalking about kernelcan i hack mz linux to limit the number of threads allowed to run?10:15
mthin theory you can hack in anything, but why do you want that feature?10:16
kristianpaulwell i want ensure better timimg response to handle IO traffic10:16
mthwouldn't thread priorities be a solution to that?10:17
kristianpaulwell i was aware of thread priorities ;)10:17
kristianpaulas i said ißm a kernel newbie10:17
mthI mean, it doesn't matter how many threads there are, it only matters that your I/O thread runs when it should10:18
kristianpaulzeah10:18
kristianpaulzes10:18
mthbut the lowest latency you'll get if you write a driver and handle interrupts10:18
mththen just put the data in a buffer and do more expensive processing from user land10:18
kristianpauli want use ram and sd as much as i can10:19
kristianpauli think NAND is slow..10:19
kristianpaulcompared to SD10:19
kristianpaulbut i dont know wnought zet#10:19
kristianpaulyet*10:19
mthyou can test by copying a fixed amount of data using "dd"10:19
mthfrom /dev/zero10:19
kristianpaulmth: i cant handle 2mhy of interrupts10:19
kristianpaulwell i must leave or plain will dropme out, cya in some hours10:20
kristianpaulthanks mth :)10:20
mthrun10:21
mthrun "sync" after "dd" and time the complete sequence10:21
mthotherwise the write cache might distort the picture10:21
mthor just copy a very large amount of data, then the cache won't have a big influence10:21
rafawpwrak: a little of headache  :)10:22
kristianpaullol10:24
kristianpaulbye10:24
wpwrakrafa: i was surprised how difficult it was for me this morning to properly rotate a 3d shape ... :) this was the orientation i was looking for http://projects.qi-hardware.com/index.php/p/ben-scans/source/tree/master/data/jpg/ben-baseframe-bottom-1mm.jpg10:25
wpwrakand this is what i got after several minutes of trying. http://projects.qi-hardware.com/index.php/p/ben-scans/source/tree/master/data/jpg/ben-baseframe-bottom-500um.jpg10:25
bartbeskristianpaul: I really only modified the sd card stuff10:26
mthAyla did a lot of fixes on gmenu2x10:32
rafawpwrak: ;-)10:45
tuxbrain2Do anyone here , lost his screen on NN? all system seems ok(checked by ssh) and it can usb.boot, I can even play music due I remember the key sequence to start blindly gmu, but screen does't turn on. it is an unopened realatively brand new NN just flashed  three or four times, the same I do the video, so it works totally ok, but I have dried the batt, and I have not touched it for two or tree days, and when I want to use it again , nor on batt, nor10:53
tuxbrain2 usb power(with or without batt) the screen is totally black10:53
mthsounds like the backlight is not working10:56
tuxbrain2any clue on how to recover it?10:57
mthcheck for broken connections, I guess10:57
mthI don't know the NN hardware very well10:57
tuxbrain2mth that was on of my guesses, but I want some assessment before... estrage due I have not messed inside with this one10:58
mthafaik it has a separate controller for the backlight10:58
tuxbrain2I must leave is some one has some clue or experience on this I will read the log.11:03
kyakhm.. i was able to create a custom font (with cyrillic glyphs) for use with setfont2.. cyrillic input/output works fine after setfont2/loadkeys/loadunimap, but the font looks terrible. Unfortunately, unclouded has been away for quite some time now, and i don't know what font/settings he used to create fonts for setfont211:39
Action: kristianpaul back11:46
kristianpaulmth: so i do dd if=/dev/zero of=foo11:47
kristianpaulfollow by sync?11:47
kristianpauloh i see numbers now11:49
kristianpaul:)11:49
mthkristianpaul: you can specify block size and block count to copy an exact amount of data13:19
wolfspraulshutting down the servers for scheduled downtime now... (server moving to new data center)16:24
wolfspraulaccording to our ISP they will be down for up to 9h, oh well16:28
wolfspraulhappy travels!16:28
wolfspraul:-)16:28
mthkristianpaul: you can use the "time" command to see how long it took22:53

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!